Empirical Evidence of Lead-Lag Relation between the Norwegian CDS and Stock Markets : Using Vector Autoregression with Exogenous Variables (VARX) and Structured Regularization for Large Vector Autoregressions with Exogenous Variables (VARX-L) Framework
MetadataShow full item record
- Master Thesis 
The master thesis studies the lead-lag relation between the Norwegian CDS and stock markets with daily observations from June 24, 2010 to May 5, 2017 of three Norwegian firms, DNB Bank ASA, Telenor ASA, and Statoil ASA. I use vector autoregression with exogenous variables models (VARX) on firm level where stock returns and credit default swap spread changes are endogenous variables, and exchange rate (NOK/Euro) change and 10-year Norwegian government bond yield change are exogenous variables. The CDS samples are drawn on senior unsecured debt with modified-modified restructuring type, Euro settlement currency, 30-year and 5-year maturity. The results of VARX suggest that the lagged equity returns predict the CDS returns while the lagged CDS spread changes do not predict the stock returns. Combined with the fact that the Norwegian stock market is more liquid than the Norwegian CDS market, one hypothesis is that CDS market is slow to reflect information due to the liquidity problem. I also use the VARX-L framework, Structured Regularization for Large Vector Autoregressions with Exogenous Variables. It implements penalty structures to the conventional VARX models. After allowing for more heterogeneity and flexible lag structure in the VARX-L models, the analysis reveals that large lags of CDS spread changes can predict stock returns. The thesis contributes to the literature in two ways. First, to the best of my knowledge, the thesis is the first to study this question in the context of Norwegian financial markets. The current literature claims that the 5-year maturity CDS is most popular and liquid contract. However, in the Norwegian CDS market, maturity seems to positively correlate with liquidity, so I also include 30-year CDS in the analysis. Secondly, the thesis adds a machine learning component to the traditional VAR analysis which allows to show that large lags of CDS spread changes can predict stock returns.